OBJECTIVES: The current study aimed to explore whether it is possible to subgroup ADHD using white matter microstructural characteristics. METHODS: In a cohort comprising subjects with ADHD (n = 227) (all aged 6-15 years) and their healthy counterparts (n = 89), the Diffusion Tensor Imaging (DTI) was used to assess fractional anisotropy (FA) in several regions: the body, genu, and splenium of corpus callosum (CC), the bilateral anterior corona radiata, the bilateral internal capsule, and the bilateral superior longitudinal fasciculus. Clinical and neuropsychological profiles were assessed using the ADHD rating scale (ADHD-RS), the Child Behavior Checklist (CBCL), the Behavior Rating Inventory of Executive Function (BRIEF), and CANTAB. Responses to methylphenidate of some of the subjects with ADHD (n = 52) were documented in the Hospital Information System. Cluster analysis was applied to subgroup the ADHD participants. Subsequent between-group comparisons were analyzed using ANCOVA and logistic regression, controlling for age and sex. RESULTS: Cluster analysis stratified the ADHD subjects into two subgroups. Subsequent analysis revealed that there are no significant differences in those behavioral measures from ADHD-RS, CBCL, or BRIEF between the two ADHD subgroups (all P >
.05). Compared with the control group, Cluster-2 exhibited lower FA and performed worse on processing speed, while Cluster-1 had higher FA but showed poorer response inhibition and sustained attention. Additionally, Cluster-2 exhibited a superior response to methylphenidate treatment compared to Cluster-1. CONCLUSIONS: Although with similar clinical features, ADHD participants could be stratified by their microstructural characteristics, which were further linked to distinct cognitive dysfunction and responses to methylphenidate.